1. Field of the Invention
The field of the invention is that of the transmission of digital data, notably in disturbed channels. More specifically, the invention relates to the transmission, in one and the same channel, of data requiring different levels of protection against transmission errors.
The transmitted data may be, for example, sound data or audiovisual data (notably in television, visiophony etc.) and, more generally, any type of digital data on which it may be worthwhile, useful or at any rate not harmful to carry out a discrimination between the digital elements using a criterion of the minimum protection level desired.
1. Description of the Prior Art
The technological background of the invention is the digital sound broadcasting system as described in the U.S. Pat. No. 4,881,241 dated 14th November 1990. The digital broadcasting system presented in these prior art patent applications is based on the joint use of a channel encoding device and a coding orthogonal frequency division multiplex (COFDM) system.
The modulation method proper of this known system consists in providing for the distribution of constituent digital elements of the data signal in the frequency-time space f-t and in the simultaneous transmission of the sets of digital elements on N parallel broadcasting channels by means of a multiplex of orthogonal carrier frequencies. This type of modulation makes it possible to prevent two successive elements of the data train from being transmitted at the same frequency. This enables the absorption of the frequency fluctuating selectivity of the channel through the frequency dispersal, during the broadcasting, of the initially adjacent digital elements.
The known encoding process is aimed, for its part, at the processing of the samples coming from the demodulator to absorb the effect of amplitude variation of the received signal, due to the Rayleigh process. The encoding is advantageously a convolutive encoding, possibly concatenated with a Reed-Solomon type of encoding.
In a known way, the encoded digital elements are furthermore interleaved, in time as well as in frequency, so as to maximize the statistical independence of the samples with respect to the Rayleigh process and to the selective character of the channel.
This method is well adapted to the broadcasting of digital signals at a high bit rate (several megabits/s) in channels that are particularly hostile. This has been demonstrated by the first embodiment or this method in digital sound radio broadcasting. In particular, it enables the reception of digital data by mobile receivers moving about in an urban environment, i.e. in the presence of parasitic noise and jamming, and under conditions of multiple propagation (Rayleigh process) generating a phenomenon of fading.
However, in its present form, this method is not used in an optimal way. The same channel encoding is used for all the data to be transmitted, with the same protection against transmission errors, irrespectively of the importance of the data elements.
It often happens that there are major differences among the pieces of digital information designed to be transmitted in the same channel. Thus, for example, in the case of sound signals, it is known that it is possible to tolerate an error rate of about 1% for the least significant bits (LSBs) while the most significant bits (MSBs) often require an error rate of less than 10.sup.-6. In the same way, in an image signal, all the transmitted coefficients do not have the same importance, especially from a psychovisual point of view.
It is clear that the error rate is related notably to the type of encoding used, all conditions of reception being moreover equal, and in particular to the error correction methods and to the redundancies introduced. It can be seen therefore that the encoding efficiency, in terms of bit rate, is related to the encoding used. In other words, the more reliable the encoding, the lower is its bit rate.
From the viewpoint of channel encoding alone, it is therefore clear that a channel encoding system that uniformly protects the flow of data and is based on the sensitivity to transmission errors of the most significant bits is sub-optimal in terms of spectral efficiency (the number of bits/s/Hz).
The result thereof is high quality encoding for all the bits, and therefore an over-coding of the bits with low significance, leading to a loss in the transmission bit rate.
There already exist known methods to match the channel encoding with the requirements of the source encoding. It has notably been proposed to use rate compatible punctured convolutional (RCPC) codes which are associated at reception, with a single Viterbi decoder working in soft decision mode. This method, described by R. V. Cox, N. Seshidri and C-E. W. Sundberg in "Combined Subband Source Coding And Convolutional Channel Coding", ITG Tagung: Digital Sprachverarbeitung, 26, Oct. 28, 1988, Bad Nauheim, achieves the periodic suppression, or puncturing, of certain bits of the source code when the maximum error rate permits it. However, this type of encoding remains related to a particular modulation, thus limiting the spectral efficiency that can be obtained. Thus, in the case of an RCPC encoding used with the 4-PSK modulation, it is possible at most to achieve a spectral efficiency that is strictly below 2. Besides it does not seem to be possible to use this technique efficiently with modulations where there are more than four phase states.
The invention is aimed at overcoming these drawbacks.